Uptime for your data
Catch data issues before your users do — Alertee runs plain SQL checks that flag silent failures: a tenant, feed, or job quietly hitting zero while everything else looks green.
Most monitors tell you something broke.
Alertee tells you what, where, and since when.
A check fires, and the alert can hand itself to your agent. It reads the check's SQL and recent history — your data, not guesswork — and tells you where the gap is and when it started.
A normal monitor
⚠ paid_payments_recorded failed
Something is wrong. Now go open a dashboard and work out what.
Alertee
Paid payments dropped to 0 at 14:05 while total payments held steady — the Stripe webhook stopped writing status = 'paid'.
What, where, and since when
SQL monitoring for the outcomes users notice
Alertee focuses on the gap between uptime and data observability: simple production data checks that keep running after setup.
Plain SQL you own
Every check is plain SQL you can read, edit, copy, or take back at any time. No black-box ML, nothing to reverse-engineer.
Multi-dimensional checks
Stratify one check by tenant, region, integration, or job — and spot the single cell failing while the rest stay green. One query, every dimension.
Schema-aware setup
Connect Postgres or ClickHouse and Alertee suggests practical SQL checks from your real schema.
Performance-aware
Read-only and light on your database — and Alertee flags slow checks so monitoring never becomes the problem it's meant to catch.
Pluggable: CLI & MCP
Plugs into your stack. Drive Alertee from your terminal or your AI assistant over MCP — and an alert can hand an agent the check to pull the SQL and recent history and investigate.
Data incident inbox
Failed data checks become owned items you can acknowledge, resolve, and classify to reduce future noise.
Run Alertee from your agent or terminal
Point Claude, Cursor, or your own agent at Alertee over MCP — it creates checks, runs them, and investigates alerts on its own. Or drive everything yourself from the CLI.
Point your client at the hosted server and sign in once — nothing to install. Pick yours:
claude mcp add --transport http alertee \
https://app.alertee.app/mcpThen run /mcp in Claude Code to sign in.
Your agent, connected to Alertee
create_check — Paid payments recorded · every 15mget_check_history — last 24hCreate & tune checks
Draft SQL checks, set schedules, snooze, and retune from chat.
Investigate alerts
An agent pulls the SQL and recent history and diagnoses on its own.
Work the inbox
Acknowledge, resolve, and triage incidents without leaving the agent.
See how setup works
From a plain-English prompt to your first alert in three steps.
Describe it
Tell Alertee what should keep working
Plain English is enough. SQL stays editable.
Suggested monitors
Generated from your schema and prompt. Edit before enabling.
Paid payments are still being recorded
SELECT count(*) FROM payments WHERE status = 'paid'
every 15m
Nightly import writes new rows
SELECT max(imported_at) FROM customer_events
daily
Every region is still reporting orders
SELECT region AS name, count(*) AS value FROM orders
every 15m
If Paid payments fails, Alertee opens one Inbox item.
Assign it, resolve it, and mark whether it was real, transient, expected, or noisy.
Simple, Transparent Pricing
Start free, then upgrade when your monitoring grows.
Every plan includes